Estimating Sample Size Requirements for Reliable Personal Authentication Using User-Specific Samples

The goal of this paper is to determine bounds for estimating minimum sample size requirement for reliable biometric identification. A new approach for the reliable estimation of the minimum sample size is proposed for arbitrary ensemble of subjects. A bound on number of acquisitions/samples per subject is arrived through an iterative procedure that tests sequences for user-specific sequences. The approach proposed in this paper is supported by information theoretic measures. These results are fundamental to the integration of concepts from statistics, complexity and probabilistic (Borel) measure spaces. We evolve a novel concept of information equivalence in comparing random sequences for its information content. Furthermore, the problem of missing or lost/corrupted matching scores is also investigated. The solution for these missing biometric matching scores is based on completeness of certain typical space and these scores can be estimated using proposed iterative algorithm.